Back to Blog
April 7, 2026 • Lucas Ropek

Anthropic Previews Powerful Mythos AI for Cybersecurity

Anthropic Previews Powerful Mythos AI for Cybersecurity Anthropic Previews Powerful Mythos AI for Cybersecurity Anthropic Previews Powerful Mythos AI for Cybersecurity

Anthropic Debuts Preview of Powerful New AI Model Mythos in Cybersecurity Initiative

Anthropic has released a preview of its new frontier model, Mythos, designed for advanced cybersecurity applications. This limited debut is part of a new initiative called Project Glasswing, involving 12 partner organizations tasked with using Mythos for defensive security work and securing critical software.

Key Features and Capabilities of Mythos

  • Powerful General-Purpose Model: Mythos is a general-purpose model within Anthropic's Claude AI systems, boasting strong agentic coding and reasoning skills. It is considered one of Anthropic's most sophisticated and high-performance models, suitable for complex tasks.
  • Cybersecurity Focus: While not exclusively trained for cybersecurity, Mythos will be employed to scan both first-party and open-source software systems for code vulnerabilities.
  • Vulnerability Detection: Anthropic claims Mythos has identified thousands of zero-day vulnerabilities, many of them critical, over recent weeks. Some of these vulnerabilities are reportedly decades old.
  • Agentic Coding and Reasoning: The model is highlighted for its advanced capabilities in coding and reasoning, making it suitable for complex agent-building tasks.

Project Glasswing and Partnerships

Project Glasswing aims to leverage Mythos for enhanced cybersecurity. The initial partner organizations include major tech players such as:

  • Amazon
  • Apple
  • Broadcom
  • Cisco
  • CrowdStrike
  • The Linux Foundation
  • Microsoft
  • Palo Alto Networks

These partners will share their findings from using Mythos, contributing to the broader tech industry's security efforts. While the preview is not generally available, 40 organizations will gain access beyond the core partnership.

Background and Previous Incidents

News of Mythos (previously codenamed "Capybara") surfaced due to a data security incident where a draft blog post was inadvertently exposed. The leaked document described Mythos as significantly more powerful than Anthropic's Opus models.

Anthropic has also faced recent security-related challenges, including the accidental exposure of source code files and the subsequent removal of thousands of GitHub repositories while attempting to rectify the issue.

Regulatory and Legal Context

Anthropic has engaged in discussions with federal officials regarding Mythos's use. However, these discussions are complicated by an ongoing legal battle between Anthropic and the Trump administration, stemming from the Pentagon labeling the AI lab a supply-chain risk.